Essay by Eric Worrall
Sadly they decided to call the project “Suncatcher” rather than “Skynet”.
Sundar Pichai says Google will start building data centers in space, powered by the sun, in 2027
- Google unveiled Project Suncatcher earlier this month.
- It aims to reduce AI’s environmental impact by relocating data centers in space, powered by the sun.
- Google CEO Sundar Pichai said the company plans to begin sending ‘machines’ to space next year.
The great AI space race has begun.
Google has been quietly working on a long-term research initiative, internally known as Project Suncatcher, to “one day scale machine learning in space.”
Google CEO Sundar Pichai told Shannon Bream on Fox News Sunday thatGoogle’s goal is to start putting data centers in space, powered by the sun, as soon as 2027.
“We are taking our first step in ’27,” he said. “We’ll send tiny, tiny racks of machines, and have them in satellites, test them out, and then start scaling from there.”
In a decade, Pichai said that it’ll be normal to build extraterrestrial data centers.
…
Google’s cosmic pivot comes amid growing global scrutiny over the power demands of data centers.
…
Read more: https://www.businessinsider.com/google-project-suncatcher-sundar-pichai-data-centers-space-solar-2027-2025-11
There might be a few issues with this plan.
Obviously there is the launch cost. Even with the insane cost of data center components, the cost of launching the whole assembly into space adds significantly to the overall cost, along with the cost of sending technicians into orbit to service the equipment.
But the biggest issue with orbital data centers might be heat dissipation. Data centers which consume hundreds of megawatts have to dissipate all of that energy as heat.
Most people think of space as cold, but for heat emitting systems like machines or humans in spacesuits, space functions more like a giant vacuum flask. The kind of vacuum which keeps your Thermos of soup hot all day on Earth completely surrounds anything you put into orbit.
On Earth, data center heat dissipation is a barely manageable problem. Data centers on Earth pump out so much heat, a sizeable portion of the cost of a constructing a data center is building the cooling system.
In space the heat dissipation problem will be far worse – you can’t just dump all the heat into a convenient large body of water located next to your data center, all the heat has to be radiated into the vacuum of space without any help from air convection.
I read the Google press release on their new orbital data center idea, and not once did they mention heat dissipation or cooling.
Why is Google advocating such unlikely sustainability ideas? First claiming data centers will be powered by nuclear fusion in the next decade, now claiming they’ll put all the data centers in space and power them with solar energy?
These grandiose plans might just be a manifestation of late stage investment bubble fever. But I wonder how bad Google’s internal divisions are, whether Google management is catching a lot of internal heat over their skyrocketing data center CO2 footprint? Perhaps these wild sustainability plans are a desperate attempt to pacify Google’s in-house climate worriers – at least until skynet suncatcher is ready to take over all the programming jobs. But that’s just my own wild theory.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Would have to calculate the economics . In a Sun synchronous orbit they could have full time sun , and their backside radiate to the dark of space .
Probably the most poorly understood essential Fact about planetary temperature is that a gray body , no matter how dark or light , comes to the same temperature as a black body . Grok does
understand this : https://x.com/i/grok/share/hdvZNCjYz0twizt9c1rH0tDna .The gray body temperature around our orbit is ~ 278.6 ( ~ 5.5c ) +- 2.3 from perihelion , coming January 3 , to aphelion in July .
The endlessly parroted 255K meme is just ” Climate Science ” hasbara .
I have in fact never seen a competent calculation of our radiative equilibrium temperature based on our actually measured spectrum ( color ) . I’d be more than happy to work with someone to do the calculations in CoSy.com .
That is not true.
A Sun-Synchronous Orbit (SSO) is a near-polar orbit where a satellite passes over the same spot on Earth on the sunlit side at roughly the same local solar time each day, maintaining consistent lighting conditions for imaging and sensing for those points along the sunlit side of its ground path. However, the same SSO satellite spends half of its orbital period traveling across the dark side of Earth.
“Typical Sun-synchronous orbits around Earth are about 600–800 km (370–500 mi) in altitude, with periods in the 96–100-minute range, and inclinations of around 98°.”
— https://en.wikipedia.org/wiki/Sun-synchronous_orbit (my bold emphasis added)
The only possible way for a SSO satellite to experience close to “full time sun” would be for its nadir ground track to be aligned with Earth’s terminator, but even then such a satellite would experience hours of eclipses at times of the spring and autumn equinoxes where it will pass through Earth’s shadow.
Wiki tells me that a radiator masses about 12 kg m-2 and will cool 100-350 W m-2. The latter figure is set by the temperature of the coolant, the hotter the better. Of course the components you are cooling don’t agree with the hotter the better. A data center rack uses at least 40 kW, modern ones up to 100 kW. Therefore the radiators will mass at least 1200 and possibly 12000 kg. For one rack.
They’ll also need batteries, the capacity of which is orbit dependent. 1 hour is a rough estimate, so 40-100 kWh, so 160-400 kg.
The rack itself will mass 1000-2000 kg
So the mass of a one rack satellite will be several tonnes, worst case almost 20 tonnes. That is feasible, it’s about 1/4 of a Skylab. For one rack.
Space junk, meteors, solar storms.
Huge amounts of data uplink and downlink.
I wonder if anyone has attempted a link margin analysis or error rate analysis.